Frequency Distribution Analysis

Analyze feature distributions, detect outliers, and explore statistical properties of model outputs

Total Features

15

3 categorical
Outliers Detected

42

2.8% of data
Skewness Range

-0.82 to 1.45

Moderate
Normality Tests

7/15

Normal dist.
Distribution Controls
Histogram with Density

Loading histogram...

Box & Violin Plot

Loading box plot...

Q-Q Normal Plot

Loading Q-Q plot...

Cumulative Distribution

Loading CDF...

Outlier Detection

Loading outliers...

Multi-Feature Distribution Comparison

Loading comparison...

Distribution Shape Analysis

Loading shape analysis...

Statistical Summary
Feature Count Mean Std Dev Min 25% Median 75% Max Skewness Kurtosis Outliers Normality
Key Insights
  • Confidence Distribution: The model shows high confidence (>0.9) for 68% of predictions, indicating good calibration
  • Outlier Analysis: Most outliers (78%) occur in low-confidence predictions, suggesting uncertainty in edge cases
  • Feature Normality: 7 out of 15 features follow normal distribution, suitable for parametric tests
  • Skewness Pattern: Positive skew in activation norms indicates presence of highly activated neurons
Recommendations
  • Address Outliers: Consider applying robust scaling or clipping to features with >5% outliers
  • Transform Skewed Features: Apply log or Box-Cox transformation to features with |skewness| > 1
  • Calibration Check: Review predictions with entropy > 0.8 for potential model uncertainty
  • Feature Engineering: Consider creating interaction terms for features with bimodal distributions